int reply_send(
struct batch_request *request) /* I (freed) */
{
#ifdef PBS_MOM
return(reply_send_mom(request));
#else
return(reply_send_svr(request));
#endif
}
static void req_stat_job_step2(
struct stat_cntl *cntl) /* I/O (free'd on return) */
{
svrattrl *pal;
job *pjob = NULL;
struct batch_request *preq;
struct batch_reply *preply;
int rc = 0;
enum TJobStatTypeEnum type;
pbs_queue *pque = NULL;
int exec_only = 0;
int bad = 0;
long DTime; /* delta time - only report full pbs_attribute list if J->MTime > DTime */
static svrattrl *dpal = NULL;
int job_array_index = 0;
job_array *pa = NULL;
char log_buf[LOCAL_LOG_BUF_SIZE];
all_jobs_iterator *iter;
preq = cntl->sc_origrq;
type = (enum TJobStatTypeEnum)cntl->sc_type;
preply = &preq->rq_reply;
/* See pbs_server_attributes(1B) for details on "poll_jobs" behaviour */
if (dpal == NULL)
{
/* build 'delta' pbs_attribute list */
svrattrl *tpal;
tlist_head dalist;
int aindex;
int atrlist[] =
{
JOB_ATR_jobname,
JOB_ATR_resc_used,
JOB_ATR_LAST
};
CLEAR_LINK(dalist);
for (aindex = 0;atrlist[aindex] != JOB_ATR_LAST;aindex++)
{
if ((tpal = attrlist_create("", "", 23)) == NULL)
{
return;
}
tpal->al_valln = atrlist[aindex];
if (dpal == NULL)
dpal = tpal;
append_link(&dalist, &tpal->al_link, tpal);
}
} /* END if (dpal == NULL) */
if (type == tjstArray)
{
pa = get_array(preq->rq_ind.rq_status.rq_id);
if (pa == NULL)
{
req_reject(PBSE_UNKARRAYID, 0, preq, NULL, "unable to find array");
return;
}
}
{
all_jobs *ajptr = NULL;
if (type == tjstQueue)
ajptr = cntl->sc_pque->qu_jobs;
else if (type == tjstSummarizeArraysQueue)
ajptr = cntl->sc_pque->qu_jobs_array_sum;
else if (type == tjstSummarizeArraysServer)
ajptr = &array_summary;
else
ajptr = &alljobs;
ajptr->lock();
iter = ajptr->get_iterator();
ajptr->unlock();
}
/*
* now ready for part 3, building the status reply,
* loop through again
*/
if ((type == tjstSummarizeArraysQueue) ||
(type == tjstSummarizeArraysServer))
{
/* No array can be owned for these options */
update_array_statuses();
}
if (type == tjstJob)
pjob = svr_find_job(preq->rq_ind.rq_status.rq_id, FALSE);
else if (type == tjstQueue)
pjob = next_job(cntl->sc_pque->qu_jobs,iter);
else if (type == tjstSummarizeArraysQueue)
pjob = next_job(cntl->sc_pque->qu_jobs_array_sum,iter);
else if (type == tjstSummarizeArraysServer)
pjob = next_job(&array_summary,iter);
else if (type == tjstArray)
{
job_array_index = -1;
pjob = NULL;
/* increment job_array_index until we find a non-null pointer or hit the end */
while (++job_array_index < pa->ai_qs.array_size)
{
if (pa->job_ids[job_array_index] != NULL)
{
if ((pjob = svr_find_job(pa->job_ids[job_array_index], FALSE)) != NULL)
{
break;
}
}
}
}
else
pjob = next_job(&alljobs,iter);
DTime = 0;
if (preq->rq_extend != NULL)
{
char *ptr;
/* FORMAT: { EXECQONLY | DELTA:<EPOCHTIME> } */
if (strstr(preq->rq_extend, EXECQUEONLY))
exec_only = 1;
ptr = strstr(preq->rq_extend, "DELTA:");
if (ptr != NULL)
{
ptr += strlen("delta:");
DTime = strtol(ptr, NULL, 10);
}
}
if ((type == tjstTruncatedServer) ||
(type == tjstTruncatedQueue))
{
long sentJobCounter;
long qjcounter;
long qmaxreport;
all_queues_iterator *iter = NULL;
svr_queues.lock();
iter = svr_queues.get_iterator();
svr_queues.unlock();
/* loop through all queues */
while ((pque = next_queue(&svr_queues,iter)) != NULL)
{
qjcounter = 0;
if ((exec_only == 1) &&
(pque->qu_qs.qu_type != QTYPE_Execution))
{
/* ignore routing queues */
unlock_queue(pque, __func__, "ignore queue", LOGLEVEL);
continue;
}
if (((pque->qu_attr[QA_ATR_MaxReport].at_flags & ATR_VFLAG_SET) != 0) &&
(pque->qu_attr[QA_ATR_MaxReport].at_val.at_long >= 0))
{
qmaxreport = pque->qu_attr[QA_ATR_MaxReport].at_val.at_long;
}
else
{
qmaxreport = TMAX_JOB;
}
if (LOGLEVEL >= 5)
{
sprintf(log_buf,"giving scheduler up to %ld idle jobs in queue %s\n",
qmaxreport,
pque->qu_qs.qu_name);
log_event(PBSEVENT_SYSTEM,PBS_EVENTCLASS_QUEUE,pque->qu_qs.qu_name,log_buf);
}
sentJobCounter = 0;
/* loop through jobs in queue */
if (pjob != NULL)
unlock_ji_mutex(pjob, __func__, "5", LOGLEVEL);
all_jobs_iterator *jobiter = NULL;
pque->qu_jobs->lock();
jobiter = pque->qu_jobs->get_iterator();
pque->qu_jobs->unlock();
while ((pjob = next_job(pque->qu_jobs,jobiter)) != NULL)
{
if ((qjcounter >= qmaxreport) &&
(pjob->ji_qs.ji_state == JOB_STATE_QUEUED))
{
/* max_report of queued jobs reached for queue */
unlock_ji_mutex(pjob, __func__, "6", LOGLEVEL);
continue;
}
pal = (svrattrl *)GET_NEXT(preq->rq_ind.rq_status.rq_attr);
rc = status_job(
pjob,
preq,
(pjob->ji_wattr[JOB_ATR_mtime].at_val.at_long >= DTime) ? pal : dpal,
&preply->brp_un.brp_status,
&bad);
if ((rc != 0) && (rc != PBSE_PERM))
{
req_reject(rc, bad, preq, NULL, NULL);
unlock_ji_mutex(pjob, __func__, "7", LOGLEVEL);
unlock_queue(pque, __func__, "perm", LOGLEVEL);
delete iter;
return;
}
sentJobCounter++;
if (pjob->ji_qs.ji_state == JOB_STATE_QUEUED)
qjcounter++;
unlock_ji_mutex(pjob, __func__, "8", LOGLEVEL);
} /* END foreach (pjob from pque) */
if (LOGLEVEL >= 5)
{
sprintf(log_buf,"sent scheduler %ld total jobs for queue %s\n",
sentJobCounter,
pque->qu_qs.qu_name);
log_event(PBSEVENT_SYSTEM,PBS_EVENTCLASS_QUEUE,pque->qu_qs.qu_name,log_buf);
}
unlock_queue(pque, __func__, "end while", LOGLEVEL);
} /* END for (pque) */
reply_send_svr(preq);
delete iter;
return;
} /* END if ((type == tjstTruncatedServer) || ...) */
while (pjob != NULL)
{
/* go ahead and build the status reply for this job */
if (exec_only)
{
if (cntl->sc_pque != NULL)
{
if (cntl->sc_pque->qu_qs.qu_type != QTYPE_Execution)
goto nextjob;
}
else
{
if (pa != NULL)
pthread_mutex_unlock(pa->ai_mutex);
pque = get_jobs_queue(&pjob);
if (pa != NULL)
pthread_mutex_lock(pa->ai_mutex);
if ((pjob == NULL) ||
(pque == NULL))
goto nextjob;
mutex_mgr pque_mutex = mutex_mgr(pque->qu_mutex, true);
if (pque->qu_qs.qu_type != QTYPE_Execution)
{
goto nextjob;
}
}
}
pal = (svrattrl *)GET_NEXT(preq->rq_ind.rq_status.rq_attr);
rc = status_job(
pjob,
preq,
pal,
&preply->brp_un.brp_status,
&bad);
if ((rc != 0) &&
(rc != PBSE_PERM))
{
if (pa != NULL)
{
unlock_ai_mutex(pa, __func__, "1", LOGLEVEL);
}
unlock_ji_mutex(pjob, __func__, "9", LOGLEVEL);
req_reject(rc, bad, preq, NULL, NULL);
delete iter;
return;
}
/* get next job */
nextjob:
if (pjob != NULL)
unlock_ji_mutex(pjob, __func__, "10", LOGLEVEL);
if (type == tjstJob)
break;
if (type == tjstQueue)
pjob = next_job(cntl->sc_pque->qu_jobs,iter);
else if (type == tjstSummarizeArraysQueue)
pjob = next_job(cntl->sc_pque->qu_jobs_array_sum,iter);
else if (type == tjstSummarizeArraysServer)
pjob = next_job(&array_summary,iter);
else if (type == tjstArray)
{
pjob = NULL;
/* increment job_array_index until we find a non-null pointer or hit the end */
while (++job_array_index < pa->ai_qs.array_size)
{
if (pa->job_ids[job_array_index] != NULL)
{
if ((pjob = svr_find_job(pa->job_ids[job_array_index], FALSE)) != NULL)
{
break;
}
}
}
}
else
pjob = next_job(&alljobs,iter);
rc = 0;
} /* END while (pjob != NULL) */
delete iter;
if (pa != NULL)
{
unlock_ai_mutex(pa, __func__, "1", LOGLEVEL);
}
reply_send_svr(preq);
if (LOGLEVEL >= 7)
{
log_event(PBSEVENT_SYSTEM,
PBS_EVENTCLASS_JOB,
"req_statjob",
"Successfully returned the status of queued jobs\n");
}
return;
} /* END req_stat_job_step2() */
int req_stat_svr(
struct batch_request *preq) /* ptr to the decoded request */
{
svrattrl *pal;
struct batch_reply *preply;
struct brp_status *pstat;
int bad = 0;
char nc_buf[128];
int numjobs;
int netrates[3];
memset(netrates, 0, sizeof(netrates));
/* update count and state counts from sv_numjobs and sv_jobstates */
lock_sv_qs_mutex(server.sv_qs_mutex, __func__);
numjobs = server.sv_qs.sv_numjobs;
unlock_sv_qs_mutex(server.sv_qs_mutex, __func__);
pthread_mutex_lock(server.sv_attr_mutex);
server.sv_attr[SRV_ATR_TotalJobs].at_val.at_long = numjobs;
server.sv_attr[SRV_ATR_TotalJobs].at_flags |= ATR_VFLAG_SET;
pthread_mutex_lock(server.sv_jobstates_mutex);
update_state_ct(
&server.sv_attr[SRV_ATR_JobsByState],
server.sv_jobstates,
server.sv_jobstbuf);
pthread_mutex_unlock(server.sv_jobstates_mutex);
netcounter_get(netrates);
snprintf(nc_buf, 127, "%d %d %d", netrates[0], netrates[1], netrates[2]);
if (server.sv_attr[SRV_ATR_NetCounter].at_val.at_str != NULL)
free(server.sv_attr[SRV_ATR_NetCounter].at_val.at_str);
server.sv_attr[SRV_ATR_NetCounter].at_val.at_str = strdup(nc_buf);
if (server.sv_attr[SRV_ATR_NetCounter].at_val.at_str != NULL)
server.sv_attr[SRV_ATR_NetCounter].at_flags |= ATR_VFLAG_SET;
pthread_mutex_unlock(server.sv_attr_mutex);
/* allocate a reply structure and a status sub-structure */
preply = &preq->rq_reply;
preply->brp_choice = BATCH_REPLY_CHOICE_Status;
CLEAR_HEAD(preply->brp_un.brp_status);
pstat = (struct brp_status *)calloc(1, sizeof(struct brp_status));
if (pstat == NULL)
{
reply_free(preply);
req_reject(PBSE_SYSTEM, 0, preq, NULL, NULL);
pthread_mutex_unlock(server.sv_attr_mutex);
return(PBSE_SYSTEM);
}
CLEAR_LINK(pstat->brp_stlink);
strcpy(pstat->brp_objname, server_name);
pstat->brp_objtype = MGR_OBJ_SERVER;
CLEAR_HEAD(pstat->brp_attr);
append_link(&preply->brp_un.brp_status, &pstat->brp_stlink, pstat);
/* add attributes to the status reply */
pal = (svrattrl *)GET_NEXT(preq->rq_ind.rq_status.rq_attr);
if (status_attrib(
pal,
svr_attr_def,
server.sv_attr,
SRV_ATR_LAST,
preq->rq_perm,
&pstat->brp_attr,
&bad,
1)) /* IsOwner == TRUE */
{
reply_badattr(PBSE_NOATTR, bad, pal, preq);
}
else
{
reply_send_svr(preq);
}
return(PBSE_NONE);
} /* END req_stat_svr() */
int req_stat_node(
struct batch_request *preq)
{
char *name;
int rc = PBSE_NONE;
int type = 0;
int bad = 0;
struct pbsnode *pnode = NULL;
struct batch_reply *preply;
struct prop props;
svrattrl *pal;
/*
* first, check that the server indeed has a list of nodes
* and if it does, validate the name of the requested object--
* either name is that of a specific node, or name[0] is null/@
* meaning request is for all nodes in the server's jurisdiction
*/
if (LOGLEVEL >= 6)
{
log_record( PBSEVENT_SCHED, PBS_EVENTCLASS_REQUEST, __func__, "entered");
}
if (svr_totnodes <= 0)
{
rc = PBSE_NONODES;
req_reject(rc, 0, preq, NULL, "node list is empty - check 'server_priv/nodes' file");
return rc;
}
name = preq->rq_ind.rq_status.rq_id;
if ((*name == '\0') || (*name == '@'))
{
type = 1;
}
else if ((*name == ':') && (*(name + 1) != '\0'))
{
if (!strcmp(name + 1, "ALL"))
{
type = 1; /* psuedo-group for all nodes */
}
else
{
type = 2;
props.name = name + 1;
props.mark = 1;
props.next = NULL;
}
}
preply = &preq->rq_reply;
preply->brp_choice = BATCH_REPLY_CHOICE_Status;
CLEAR_HEAD(preply->brp_un.brp_status);
if (type == 0)
{
/* get status of the named node */
pnode = find_nodebyname(name);
if (pnode == NULL)
{
rc = PBSE_UNKNODE;
req_reject(rc, 0, preq, NULL, "cannot locate specified node");
return(rc);
}
/* get the status on all of the numa nodes */
if (pnode->nd_is_alps_reporter == TRUE)
rc = get_alps_statuses(pnode, preq, &bad, &preply->brp_un.brp_status);
else
rc = get_numa_statuses(pnode, preq, &bad, &preply->brp_un.brp_status);
unlock_node(pnode, __func__, "type == 0", LOGLEVEL);
}
else
{
/* get status of all or several nodes */
all_nodes_iterator *iter = NULL;
while ((pnode = next_host(&allnodes,&iter,NULL)) != NULL)
{
if ((type == 2) &&
(!hasprop(pnode, &props)))
{
unlock_node(pnode, __func__, "type != 0, next_host", LOGLEVEL);
continue;
}
/* get the status on all of the numa nodes */
if (pnode->nd_is_alps_reporter == TRUE)
rc = get_alps_statuses(pnode, preq, &bad, &preply->brp_un.brp_status);
else
rc = get_numa_statuses(pnode, preq, &bad, &preply->brp_un.brp_status);
if (rc != PBSE_NONE)
{
unlock_node(pnode, __func__, "type != 0, rc != 0, get_numa_statuses", LOGLEVEL);
break;
}
unlock_node(pnode, __func__, "type != 0, rc == 0, get_numa_statuses", LOGLEVEL);
}
if (iter != NULL)
delete iter;
}
if (rc == PBSE_NONE)
{
/* SUCCESS */
reply_send_svr(preq);
}
else
{
if (rc != PBSE_UNKNODEATR)
{
req_reject(rc, 0, preq, NULL, NULL);
}
else
{
pal = (svrattrl *)GET_NEXT(preq->rq_ind.rq_status.rq_attr);
reply_badattr(rc, bad, pal, preq);
}
}
return(rc);
} /* END req_stat_node() */
int req_stat_que(
batch_request *preq)
{
char *name;
pbs_queue *pque = NULL;
struct batch_reply *preply;
int rc = 0;
int type = 0;
char log_buf[LOCAL_LOG_BUF_SIZE+1];
/*
* first, validate the name of the requested object, either
* a queue, or null for all queues
*/
name = preq->rq_ind.rq_status.rq_id;
if ((*name == '\0') || (*name == '@'))
{
type = 1;
}
else
{
pque = find_queuebyname(name);
if (pque == NULL)
{
rc = PBSE_UNKQUE;
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "cannot locate queue %s", name);
req_reject(rc, 0, preq, NULL, log_buf);
return rc;
}
}
preply = &preq->rq_reply;
preply->brp_choice = BATCH_REPLY_CHOICE_Status;
CLEAR_HEAD(preply->brp_un.brp_status);
if (type == 0)
{
/* get status of the named queue */
mutex_mgr pque_mutex = mutex_mgr(pque->qu_mutex, true);
rc = status_que(pque, preq, &preply->brp_un.brp_status);
/* pque_qu_mutex will be unlocked in the destructor when we leave this scope */
}
else
{
/* pque == NULL before next_queue */
svr_queues.lock();
all_queues_iterator *iter = svr_queues.get_iterator();
svr_queues.unlock();
/* get status of all queues */
while ((pque = next_queue(&svr_queues,iter)) != NULL)
{
mutex_mgr pque_mutex = mutex_mgr(pque->qu_mutex, true);
rc = status_que(pque, preq, &preply->brp_un.brp_status);
if (rc != 0)
{
if (rc != PBSE_PERM)
{
break;
}
rc = 0;
}
}
delete iter;
}
if (rc != PBSE_NONE)
{
reply_free(preply);
req_reject(PBSE_NOATTR, rc, preq, NULL, "status_queue failed");
}
else
{
reply_send_svr(preq);
}
return rc;
} /* END req_stat_que() */
static void req_stat_job_step2(
struct stat_cntl *cntl) /* I/O (free'd on return) */
{
svrattrl *pal;
job *pjob = NULL;
struct batch_request *preq;
struct batch_reply *preply;
int rc = 0;
enum TJobStatTypeEnum type;
pbs_queue *pque = NULL;
int exec_only = 0;
int bad = 0;
long DTime; /* delta time - only report full pbs_attribute list if J->MTime > DTime */
static svrattrl *dpal = NULL;
int job_array_index = 0;
job_array *pa = NULL;
char log_buf[LOCAL_LOG_BUF_SIZE];
int iter;
time_t time_now = time(NULL);
long poll_jobs = 0;
char job_id[PBS_MAXSVRJOBID+1];
int job_substate = -1;
time_t job_momstattime = -1;
preq = cntl->sc_origrq;
type = (enum TJobStatTypeEnum)cntl->sc_type;
preply = &preq->rq_reply;
/* See pbs_server_attributes(1B) for details on "poll_jobs" behaviour */
if (dpal == NULL)
{
/* build 'delta' pbs_attribute list */
svrattrl *tpal;
tlist_head dalist;
int aindex;
int atrlist[] =
{
JOB_ATR_jobname,
JOB_ATR_resc_used,
JOB_ATR_LAST
};
CLEAR_LINK(dalist);
for (aindex = 0;atrlist[aindex] != JOB_ATR_LAST;aindex++)
{
if ((tpal = attrlist_create("", "", 23)) == NULL)
{
return;
}
tpal->al_valln = atrlist[aindex];
if (dpal == NULL)
dpal = tpal;
append_link(&dalist, &tpal->al_link, tpal);
}
} /* END if (dpal == NULL) */
if (type == tjstArray)
{
pa = get_array(preq->rq_ind.rq_status.rq_id);
if (pa == NULL)
{
req_reject(PBSE_UNKARRAYID, 0, preq, NULL, "unable to find array");
return;
}
}
iter = -1;
get_svr_attr_l(SRV_ATR_PollJobs, &poll_jobs);
if (!poll_jobs)
{
/* polljobs not set - indicates we may need to obtain fresh data from
MOM */
if (cntl->sc_jobid[0] == '\0')
pjob = NULL;
else
pjob = svr_find_job(cntl->sc_jobid, FALSE);
while (1)
{
if (pjob == NULL)
{
/* start from the first job */
if (type == tjstJob)
{
pjob = svr_find_job(preq->rq_ind.rq_status.rq_id, FALSE);
}
else if (type == tjstQueue)
{
pjob = next_job(cntl->sc_pque->qu_jobs,&iter);
}
else if (type == tjstArray)
{
job_array_index = 0;
/* increment job_array_index until we find a non-null pointer or hit the end */
while (job_array_index < pa->ai_qs.array_size)
{
if (pa->job_ids[job_array_index] != NULL)
{
if ((pjob = svr_find_job(pa->job_ids[job_array_index], FALSE)) != NULL)
{
unlock_ji_mutex(pjob, __func__, "2", LOGLEVEL);
break;
}
}
job_array_index++;
}
}
else
{
pjob = next_job(&alljobs,&iter);
}
} /* END if (pjob == NULL) */
else
{
strcpy(job_id, pjob->ji_qs.ji_jobid);
unlock_ji_mutex(pjob, __func__, "3", LOGLEVEL);
if (type == tjstJob)
break;
if (type == tjstQueue)
pjob = next_job(cntl->sc_pque->qu_jobs,&iter);
else if (type == tjstArray)
{
pjob = NULL;
/* increment job_array_index until we find a non-null pointer or hit the end */
while (++job_array_index < pa->ai_qs.array_size)
{
if (pa->job_ids[job_array_index] != NULL)
{
if ((pjob = svr_find_job(pa->job_ids[job_array_index], FALSE)) != NULL)
{
unlock_ji_mutex(pjob, __func__, "3", LOGLEVEL);
break;
}
}
}
}
else
pjob = next_job(&alljobs,&iter);
}
if (pjob == NULL)
break;
strcpy(job_id, pjob->ji_qs.ji_jobid);
job_substate = pjob->ji_qs.ji_substate;
job_momstattime = pjob->ji_momstat;
strcpy(cntl->sc_jobid, job_id);
unlock_ji_mutex(pjob, __func__, "4", LOGLEVEL);
pjob = NULL;
/* PBS_RESTAT_JOB defaults to 30 seconds */
if ((job_substate == JOB_SUBSTATE_RUNNING) &&
((time_now - job_momstattime) > JobStatRate))
{
/* go to MOM for status */
if ((rc = stat_to_mom(job_id, cntl)) == PBSE_MEM_MALLOC)
break;
if (rc != 0)
{
pjob = svr_find_job(job_id, FALSE);
rc = 0;
continue;
}
if (pa != NULL)
unlock_ai_mutex(pa, __func__, "1", LOGLEVEL);
return; /* will pick up after mom replies */
}
} /* END while(1) */
if (rc != 0)
{
if (pa != NULL)
unlock_ai_mutex(pa, __func__, "2", LOGLEVEL);
reply_free(preply);
req_reject(rc, 0, preq, NULL, "cannot get update from mom");
return;
}
} /* END if (!server.sv_attr[SRV_ATR_PollJobs].at_val.at_long) */
/*
* now ready for part 3, building the status reply,
* loop through again
*/
if ((type == tjstSummarizeArraysQueue) ||
(type == tjstSummarizeArraysServer))
{
/* No array can be owned for these options */
update_array_statuses();
}
if (type == tjstJob)
pjob = svr_find_job(preq->rq_ind.rq_status.rq_id, FALSE);
else if (type == tjstQueue)
pjob = next_job(cntl->sc_pque->qu_jobs,&iter);
else if (type == tjstSummarizeArraysQueue)
pjob = next_job(cntl->sc_pque->qu_jobs_array_sum,&iter);
else if (type == tjstSummarizeArraysServer)
pjob = next_job(&array_summary,&iter);
else if (type == tjstArray)
{
job_array_index = -1;
pjob = NULL;
/* increment job_array_index until we find a non-null pointer or hit the end */
while (++job_array_index < pa->ai_qs.array_size)
{
if (pa->job_ids[job_array_index] != NULL)
{
if ((pjob = svr_find_job(pa->job_ids[job_array_index], FALSE)) != NULL)
{
break;
}
}
}
}
else
pjob = next_job(&alljobs,&iter);
DTime = 0;
if (preq->rq_extend != NULL)
{
char *ptr;
/* FORMAT: { EXECQONLY | DELTA:<EPOCHTIME> } */
if (strstr(preq->rq_extend, EXECQUEONLY))
exec_only = 1;
ptr = strstr(preq->rq_extend, "DELTA:");
if (ptr != NULL)
{
ptr += strlen("delta:");
DTime = strtol(ptr, NULL, 10);
}
}
if ((type == tjstTruncatedServer) ||
(type == tjstTruncatedQueue))
{
long sentJobCounter;
long qjcounter;
long qmaxreport;
int iter = -1;
/* loop through all queues */
while ((pque = next_queue(&svr_queues,&iter)) != NULL)
{
qjcounter = 0;
if ((exec_only == 1) &&
(pque->qu_qs.qu_type != QTYPE_Execution))
{
/* ignore routing queues */
unlock_queue(pque, __func__, "ignore queue", LOGLEVEL);
continue;
}
if (((pque->qu_attr[QA_ATR_MaxReport].at_flags & ATR_VFLAG_SET) != 0) &&
(pque->qu_attr[QA_ATR_MaxReport].at_val.at_long >= 0))
{
qmaxreport = pque->qu_attr[QA_ATR_MaxReport].at_val.at_long;
}
else
{
qmaxreport = TMAX_JOB;
}
if (LOGLEVEL >= 5)
{
sprintf(log_buf,"giving scheduler up to %ld idle jobs in queue %s\n",
qmaxreport,
pque->qu_qs.qu_name);
log_event(PBSEVENT_SYSTEM,PBS_EVENTCLASS_QUEUE,pque->qu_qs.qu_name,log_buf);
}
sentJobCounter = 0;
/* loop through jobs in queue */
if (pjob != NULL)
unlock_ji_mutex(pjob, __func__, "5", LOGLEVEL);
iter = -1;
while ((pjob = next_job(pque->qu_jobs,&iter)) != NULL)
{
if ((qjcounter >= qmaxreport) &&
(pjob->ji_qs.ji_state == JOB_STATE_QUEUED))
{
/* max_report of queued jobs reached for queue */
unlock_ji_mutex(pjob, __func__, "6", LOGLEVEL);
continue;
}
pal = (svrattrl *)GET_NEXT(preq->rq_ind.rq_status.rq_attr);
rc = status_job(
pjob,
preq,
(pjob->ji_wattr[JOB_ATR_mtime].at_val.at_long >= DTime) ? pal : dpal,
&preply->brp_un.brp_status,
&bad);
if ((rc != 0) && (rc != PBSE_PERM))
{
req_reject(rc, bad, preq, NULL, NULL);
if (pa != NULL)
{
unlock_ai_mutex(pa, __func__, "1", LOGLEVEL);
}
unlock_ji_mutex(pjob, __func__, "7", LOGLEVEL);
unlock_queue(pque, __func__, "perm", LOGLEVEL);
return;
}
sentJobCounter++;
if (pjob->ji_qs.ji_state == JOB_STATE_QUEUED)
qjcounter++;
unlock_ji_mutex(pjob, __func__, "8", LOGLEVEL);
} /* END foreach (pjob from pque) */
if (LOGLEVEL >= 5)
{
sprintf(log_buf,"sent scheduler %ld total jobs for queue %s\n",
sentJobCounter,
pque->qu_qs.qu_name);
log_event(PBSEVENT_SYSTEM,PBS_EVENTCLASS_QUEUE,pque->qu_qs.qu_name,log_buf);
}
unlock_queue(pque, __func__, "end while", LOGLEVEL);
} /* END for (pque) */
if (pa != NULL)
unlock_ai_mutex(pa, __func__, "1", LOGLEVEL);
reply_send_svr(preq);
return;
} /* END if ((type == tjstTruncatedServer) || ...) */
while (pjob != NULL)
{
/* go ahead and build the status reply for this job */
if (exec_only)
{
if (cntl->sc_pque != NULL)
{
if (cntl->sc_pque->qu_qs.qu_type != QTYPE_Execution)
goto nextjob;
}
else
{
if (pa != NULL)
pthread_mutex_unlock(pa->ai_mutex);
pque = get_jobs_queue(&pjob);
if (pa != NULL)
pthread_mutex_lock(pa->ai_mutex);
if ((pjob == NULL) ||
(pque == NULL))
goto nextjob;
if (pque->qu_qs.qu_type != QTYPE_Execution)
{
unlock_queue(pque, __func__, "not exec", LOGLEVEL);
goto nextjob;
}
unlock_queue(pque, __func__, "exec", LOGLEVEL);
}
}
pal = (svrattrl *)GET_NEXT(preq->rq_ind.rq_status.rq_attr);
rc = status_job(
pjob,
preq,
pal,
&preply->brp_un.brp_status,
&bad);
if ((rc != 0) &&
(rc != PBSE_PERM))
{
if (pa != NULL)
{
unlock_ai_mutex(pa, __func__, "1", LOGLEVEL);
}
unlock_ji_mutex(pjob, __func__, "9", LOGLEVEL);
req_reject(rc, bad, preq, NULL, NULL);
return;
}
/* get next job */
nextjob:
if (pjob != NULL)
unlock_ji_mutex(pjob, __func__, "10", LOGLEVEL);
if (type == tjstJob)
break;
if (type == tjstQueue)
pjob = next_job(cntl->sc_pque->qu_jobs,&iter);
else if (type == tjstSummarizeArraysQueue)
pjob = next_job(cntl->sc_pque->qu_jobs_array_sum,&iter);
else if (type == tjstSummarizeArraysServer)
pjob = next_job(&array_summary,&iter);
else if (type == tjstArray)
{
pjob = NULL;
/* increment job_array_index until we find a non-null pointer or hit the end */
while (++job_array_index < pa->ai_qs.array_size)
{
if (pa->job_ids[job_array_index] != NULL)
{
if ((pjob = svr_find_job(pa->job_ids[job_array_index], FALSE)) != NULL)
{
break;
}
}
}
}
else
pjob = next_job(&alljobs,&iter);
rc = 0;
} /* END while (pjob != NULL) */
if (pa != NULL)
{
unlock_ai_mutex(pa, __func__, "1", LOGLEVEL);
}
reply_send_svr(preq);
if (LOGLEVEL >= 7)
{
log_event(PBSEVENT_SYSTEM,
PBS_EVENTCLASS_JOB,
"req_statjob",
"Successfully returned the status of queued jobs\n");
}
return;
} /* END req_stat_job_step2() */
int req_stat_que(
struct batch_request *preq)
{
char *name;
pbs_queue *pque = NULL;
struct batch_reply *preply;
int rc = 0;
int type = 0;
char log_buf[LOCAL_LOG_BUF_SIZE+1];
/*
* first, validate the name of the requested object, either
* a queue, or null for all queues
*/
name = preq->rq_ind.rq_status.rq_id;
if ((*name == '\0') || (*name == '@'))
{
type = 1;
}
else
{
pque = find_queuebyname(name);
if (pque == NULL)
{
rc = PBSE_UNKQUE;
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "can not locate queue %s", name);
req_reject(rc, 0, preq, NULL, log_buf);
return rc;
}
}
preply = &preq->rq_reply;
preply->brp_choice = BATCH_REPLY_CHOICE_Status;
CLEAR_HEAD(preply->brp_un.brp_status);
if (type == 0)
{
/* get status of the named queue */
rc = status_que(pque, preq, &preply->brp_un.brp_status);
unlock_queue(pque, "req_stat_que", "type == 0", LOGLEVEL);
}
else
{
/* pque == NULL before next_queue */
int iter = -1;
/* get status of all queues */
while ((pque = next_queue(&svr_queues,&iter)) != NULL)
{
rc = status_que(pque, preq, &preply->brp_un.brp_status);
if (rc != 0)
{
if (rc != PBSE_PERM)
{
unlock_queue(pque, "req_stat_que", "break", LOGLEVEL);
break;
}
rc = 0;
}
unlock_queue(pque, "req_stat_que", "end while", LOGLEVEL);
}
}
if (rc != PBSE_NONE)
{
reply_free(preply);
req_reject(PBSE_NOATTR, rc, preq, NULL, "status_queue failed");
}
else
{
reply_send_svr(preq);
}
return rc;
} /* END req_stat_que() */
int req_locatejob(
struct batch_request *preq)
{
int rc = PBSE_NONE;
char *at;
int i;
job *pjob;
char *location = (char *)0;
if ((at = strchr(preq->rq_ind.rq_locate, (int)'@')))
* at = '\0'; /* strip off @server_name */
pjob = svr_find_job(preq->rq_ind.rq_locate, FALSE);
if (pjob)
{
unlock_ji_mutex(pjob, __func__, "1", LOGLEVEL);
location = server_name;
}
else
{
for (i = 0; i < server.sv_tracksize; i++)
{
if ((server.sv_track + i)->tk_mtime &&
!strcmp((server.sv_track + i)->tk_jobid, preq->rq_ind.rq_locate))
{
location = (server.sv_track + i)->tk_location;
break;
}
}
}
if (location != NULL)
{
preq->rq_reply.brp_code = 0;
preq->rq_reply.brp_auxcode = 0;
preq->rq_reply.brp_choice = BATCH_REPLY_CHOICE_Locate;
snprintf(preq->rq_reply.brp_un.brp_locate, sizeof(preq->rq_reply.brp_un.brp_locate), "%s", location);
reply_send_svr(preq);
}
else
{
if (LOGLEVEL >= 7)
{
log_event(
PBSEVENT_JOB,
PBS_EVENTCLASS_JOB,
preq->rq_ind.rq_locate,
"cannot find job in server tracking list");
}
rc = PBSE_UNKJOBID;
req_reject(rc, 0, preq, NULL, NULL);
}
return rc;
} /* END req_locatejob() */
void req_stat_job_step2(
struct stat_cntl *cntl) /* I/O (free'd on return) */
{
batch_request *preq = cntl->sc_origrq;
svrattrl *pal = (svrattrl *)GET_NEXT(preq->rq_ind.rq_status.rq_attr);
job *pjob = NULL;
struct batch_reply *preply = &preq->rq_reply;
int rc = 0;
enum TJobStatTypeEnum type = (enum TJobStatTypeEnum)cntl->sc_type;
bool exec_only = false;
int bad = 0;
/* delta time - only report full pbs_attribute list if J->MTime > DTime */
int job_array_index = -1;
job_array *pa = NULL;
all_jobs_iterator *iter;
if (preq->rq_extend != NULL)
{
/* FORMAT: { EXECQONLY } */
if (strstr(preq->rq_extend, EXECQUEONLY))
exec_only = true;
}
if ((type == tjstTruncatedServer) ||
(type == tjstTruncatedQueue))
{
handle_truncated_qstat(exec_only, cntl->sc_condensed, preq);
return;
} /* END if ((type == tjstTruncatedServer) || ...) */
else if (type == tjstJob)
{
pjob = svr_find_job(preq->rq_ind.rq_status.rq_id, FALSE);
if (pjob != NULL)
{
if ((rc = status_job(pjob, preq, pal, &preply->brp_un.brp_status, cntl->sc_condensed, &bad)))
req_reject(rc, bad, preq, NULL, NULL);
else
reply_send_svr(preq);
unlock_ji_mutex(pjob, __func__, "1", LOGLEVEL);
}
else
{
req_reject(PBSE_JOBNOTFOUND, bad, preq, NULL, NULL);
}
}
else
{
if (type == tjstArray)
{
pa = get_array(preq->rq_ind.rq_status.rq_id);
if (pa == NULL)
{
req_reject(PBSE_UNKARRAYID, 0, preq, NULL, "unable to find array");
return;
}
}
else if ((type == tjstSummarizeArraysQueue) ||
(type == tjstSummarizeArraysServer))
update_array_statuses();
iter = get_correct_status_iterator(cntl);
for (pjob = get_next_status_job(cntl, job_array_index, pa, iter);
pjob != NULL;
pjob = get_next_status_job(cntl, job_array_index, pa, iter))
{
mutex_mgr job_mutex(pjob->ji_mutex, true);
/* go ahead and build the status reply for this job */
if (pjob->ji_being_recycled == true)
continue;
if (exec_only)
{
if (cntl->sc_pque != NULL)
{
if (cntl->sc_pque->qu_qs.qu_type != QTYPE_Execution)
continue;
}
else if (in_execution_queue(pjob, pa) == false)
continue;
}
rc = status_job(pjob, preq, pal, &preply->brp_un.brp_status, cntl->sc_condensed, &bad);
if ((rc != PBSE_NONE) &&
(rc != PBSE_PERM))
{
if (pa != NULL)
unlock_ai_mutex(pa, __func__, "1", LOGLEVEL);
req_reject(rc, bad, preq, NULL, NULL);
delete iter;
return;
}
} /* END for (pjob != NULL) */
delete iter;
if (pa != NULL)
{
unlock_ai_mutex(pa, __func__, "1", LOGLEVEL);
}
reply_send_svr(preq);
}
if (LOGLEVEL >= 7)
{
log_event(PBSEVENT_SYSTEM,
PBS_EVENTCLASS_JOB,
"req_statjob",
"Successfully returned the status of queued jobs\n");
}
return;
} /* END req_stat_job_step2() */
void handle_truncated_qstat(
bool exec_only,
bool condensed,
batch_request *preq)
{
long sentJobCounter = 0;
long qmaxreport;
all_queues_iterator *queue_iter = NULL;
pbs_queue *pque;
char log_buf[LOCAL_LOG_BUF_SIZE];
job *pjob;
svrattrl *pal = (svrattrl *)GET_NEXT(preq->rq_ind.rq_status.rq_attr);
batch_reply *preply = &preq->rq_reply;
int bad = 0;
svr_queues.lock();
queue_iter = svr_queues.get_iterator();
svr_queues.unlock();
/* loop through all queues */
while ((pque = next_queue(&svr_queues, queue_iter)) != NULL)
{
long qjcounter = 0;
mutex_mgr queue_mutex(pque->qu_mutex, true);
if ((exec_only == true) &&
(pque->qu_qs.qu_type != QTYPE_Execution))
{
/* ignore routing queues */
continue;
}
if (((pque->qu_attr[QA_ATR_MaxReport].at_flags & ATR_VFLAG_SET) != 0) &&
(pque->qu_attr[QA_ATR_MaxReport].at_val.at_long >= 0))
{
qmaxreport = pque->qu_attr[QA_ATR_MaxReport].at_val.at_long;
}
else
{
qmaxreport = TMAX_JOB;
}
if (LOGLEVEL >= 5)
{
snprintf(log_buf, sizeof(log_buf), "Reporting up to %ld idle jobs in queue %s\n",
qmaxreport,
pque->qu_qs.qu_name);
log_event(PBSEVENT_SYSTEM,PBS_EVENTCLASS_QUEUE,pque->qu_qs.qu_name,log_buf);
}
/* loop through jobs in queue */
all_jobs_iterator *jobiter = NULL;
pque->qu_jobs->lock();
jobiter = pque->qu_jobs->get_iterator();
pque->qu_jobs->unlock();
while ((pjob = next_job(pque->qu_jobs, jobiter)) != NULL)
{
mutex_mgr job_mgr(pjob->ji_mutex, true);
if ((qjcounter >= qmaxreport) &&
(pjob->ji_qs.ji_state == JOB_STATE_QUEUED))
{
/* max_report of queued jobs reached for queue */
continue;
}
int rc = status_job(pjob, preq, pal, &preply->brp_un.brp_status, condensed, &bad);
if ((rc != 0) &&
(rc != PBSE_PERM))
{
req_reject(rc, bad, preq, NULL, NULL);
delete queue_iter;
return;
}
sentJobCounter++;
if (pjob->ji_qs.ji_state == JOB_STATE_QUEUED)
qjcounter++;
} /* END foreach (pjob from pque) */
if (LOGLEVEL >= 5)
{
snprintf(log_buf, sizeof(log_buf), "Reported %ld total jobs for queue %s\n",
sentJobCounter,
pque->qu_qs.qu_name);
log_event(PBSEVENT_SYSTEM,PBS_EVENTCLASS_QUEUE,pque->qu_qs.qu_name,log_buf);
}
} /* END for (pque) */
reply_send_svr(preq);
delete queue_iter;
return;
} // END handle_truncated_qstat()
